Using medication administration and double-check data to infer social network influence in intensive care units

Identifying individuals with positive patient safety influence can be challenging and often involve time-consuming surveys. However, as healthcare systems increase their integration of health information technologies with clinical workflow, more social and behavioral data can be obtained automatically. Physician referrals, surgical team schedules, and medication double-check logs are examples of behavioral data that reflect a degree of individual choice, preference, and influence. In this paper, we explored data from medication double-check interactions in the context of nursing influence. Survey responses and double-check data from 86 nurses in three intensive care units were collected. There was on average 21.8% overlap between influence and double-check networks. Medication in-degree scores tended to correlate more with influence, especially when it came to trust proxies for two units (0.79 and 0.47). While results varied by unit and medication class, this work provides a lens for exploring influence on units using unobtrusive measures.